Contrast-independent partially explicit time discretizations for multiscale wave problems

Eric T. Chung, Yalchin Efendiev, Wing Tat Leung, Petr N. Vabishchevich

TL;DR

This paper introduces a novel "contrast-independent partially explicit time discretization" method for simulating wave propagation in complex, heterogeneous materials. By cleverly splitting the problem into "fast" and "slow" components and treating them with different numerical approaches, the method achieves stable simulations with time steps that are independent of the material's extreme property variations, a significant improvement over existing explicit methods.

What Problem Does This Paper Address?

Simulating wave phenomena in materials with highly contrasting properties (e.g., materials with tiny, highly conductive channels embedded in a much larger, less conductive matrix) is a major challenge. Traditional explicit numerical methods, while often preferred for their simplicity and speed, require extremely small time steps to remain stable when dealing with these high contrasts. This makes simulations computationally prohibitive. Implicit methods can handle these contrasts but come with a higher computational cost per time step. The authors aim to bridge this gap by developing a method that offers the stability of implicit methods with a time step independent of the contrast, while retaining some of the efficiency of explicit methods.

Methods

The core of the paper is a novel temporal splitting technique. They decompose the spatial domain into two subspaces: one that captures the "contrast-dependent" (fast) features and another for the "contrast-independent" (slow) features. This decomposition is inspired by multiscale methods, particularly the Generalized Multiscale Finite Element Method (GMsFEM).

The key idea is to treat the "slow" components implicitly (or with a stable scheme) and the "fast" components explicitly. This is achieved by splitting the wave equation into two coupled equations, each operating on a different subspace. The paper rigorously analyzes the stability of this splitting, proving unconditional stability under certain conditions for the "slow" subspace. They also explore different ways to construct these subspaces, including using constrained energy minimization (CEM) techniques.

Datasets

Not discussed in the paper. The paper focuses on developing and analyzing a numerical method, and the numerical examples use synthetic heterogeneous media with varying contrast and multiscale features.

Key Insights and Ideas

  • Contrast Independence is Key: The primary goal is to decouple the time-stepping stability from the material contrast. This is crucial for practical simulations of real-world heterogeneous media.
  • Spatial Decomposition as a Tool: The success of the method hinges on an effective spatial decomposition that separates the "fast" (high contrast) and "slow" (low contrast) physics. This allows for targeted numerical treatment.
  • Hybrid Explicit-Implicit Approach: By treating different parts of the solution with different numerical schemes (implicit for slow, explicit for fast), they achieve a balance between stability and computational cost.
  • Stability Guarantees: The authors provide theoretical stability analysis, demonstrating that their proposed scheme can be unconditionally stable under specific conditions, which is a significant theoretical contribution.
  • GMsFEM as a Foundation: The method builds upon existing multiscale techniques like GMsFEM, leveraging its ability to capture multiscale features and construct appropriate basis functions.

Key Results

The paper presents numerical results that demonstrate the effectiveness of their proposed method.

  • Similar Accuracy to Implicit Methods: The numerical examples show that the partially explicit method achieves accuracy comparable to fully implicit methods.
  • Contrast-Independent Time Steps: The method allows for time steps that are independent of the material contrast, a significant advantage over standard explicit methods.
  • Improved Results with Additional Basis Functions: The use of additional basis functions (derived from GMsFEM) significantly improves the accuracy of the simulations, especially for challenging source terms and media.
  • Performance Comparison: The proposed method performs similarly to implicit methods that also use additional basis functions, suggesting that treating these additional degrees of freedom explicitly is a viable strategy.

Limitations and Future Work

  • Mass Matrix Coupling: The proposed method is still coupled via the mass matrix. While they briefly discuss mass lumping as a way to remove this coupling, its performance was not optimal in their tests.
  • Construction of Subspaces: The effectiveness of the method relies on the proper construction of the "fast" and "slow" spatial subspaces. The paper explores some options but suggests further research into optimal subspace construction.
  • Mass Lumping Performance: The authors note that mass lumping, while a potential avenue for improvement, requires further investigation with other discretization techniques (e.g., mixed or discontinuous Galerkin methods) for better performance.
  • Generalization: While the paper focuses on wave equations, the underlying principles of contrast-independent splitting could potentially be extended to other multiscale problems.

Summary updated

Abstract

In this work, we design and investigate contrast-independent partially explicit time discretizations for wave equations in heterogeneous high-contrast media. We consider multiscale problems, where the spatial heterogeneities are at subgrid level and are not resolved. In our previous work, we have introduced contrast-independent partially explicit time discretizations and applied to parabolic equations. The main idea of contrast-independent partially explicit time discretization is to split the spatial space into two components: contrast dependent (fast) and contrast independent (slow) spaces defined via multiscale space decomposition. Using this decomposition, our goal is further appropriately to introduce time splitting such that the resulting scheme is stable and can guarantee contrast-independent discretization under some suitable (reasonable) conditions. In this paper, we propose contrast-independent partially explicitly scheme for wave equations. The splitting requires a careful design. We prove that the proposed splitting is unconditionally stable under some suitable conditions formulated for the second space (slow). This condition requires some type of non-contrast dependent space and is easier to satisfy in the "slow" space. We present numerical results and show that the proposed methods provide results similar to implicit methods with the time step that is independent of the contrast.
BibTeX
@article{chung2022contrast,
  author = {Eric T. Chung and Yalchin Efendiev and Wing Tat Leung and Petr N. Vabishchevich},
  title = {{Contrast-independent partially explicit time discretizations for multiscale wave problems}},
  journal = {Journal of Computational Physics},
  year = {2022},
  doi = {10.1016/j.jcp.2022.111226},
}
Figure